AUTOMATIC ANALYSIS OF MAGNETOGRAM SEQUENCES FOR SOLAR FLARES FORECASTING

Authors

  • Luis Fernando L. Grim Faculdade de Tecnologia – Universidade Estadual de Campinas; Instituto Federal de Sao Paulo
  • Andre Leon S. Gradvohl Faculdade de Tecnologia – Universidade Estadual de Campinas

Keywords:

Solar Flares, Deep Learning, Magnetogram Sequences, Forecasting

Abstract

Solar flares are releases of electromagnetic energy that occur on the Sun’s surface and can reach the Earth’s atmosphere in a few hours or minutes. Depending on the flares intensity, such as M- or X-class flares, they can have a big impact on some activities and technologies, particularly satellite, telecommunications, and electrical power systems. Therefore, predicting the occurrence of solar flares helps mitigate their effects on Earth. Solar flares occur in active solar regions. These regions with magnetic fields, known as sunspots, can be precursors to solar flares. Observing the evolution of active regions through a set of solar images, such as magnetograms, is possible. A magnetogram is an image of the intensity and location of magnetic fields in the solar magnetosphere. The emergence process of solar flares generally lasts from 72h to 96h. Thus, more recent works started to combine image-based and time series Deep Learning models to consider the evolution of active regions in the Sun. Therefore, we propose an automatic solar flare forecasting system based on 3D Convolutional Neural Networks, capable of extracting learning parameters in the temporal spectrum of an image sequence, for forecasting an M or X classes solar flare in the next 48h. In preliminary tests, the fine-tuned ResNet 2D+1 model reached compatible results (F1 = 0.56, HSS = 0.55, TSS = 0.48) according to the state-of-the-art models optimized for the F1 metric.

Downloads

Published

2022-12-10

How to Cite

Grim, L. F. L., & Gradvohl, A. L. S. (2022). AUTOMATIC ANALYSIS OF MAGNETOGRAM SEQUENCES FOR SOLAR FLARES FORECASTING. Journal of Production and Automation (JPAUT) ISSN 2595-9573, 5(2), 2–9. Retrieved from https://jpaut.com.br/index.php/jpaut/article/view/2